What do you do when you brief a presidential administration on your latest AI model, only to have that same administration publicly declare your relationship over a week later?
This is the awkward position Anthropic finds itself in after co-founder Jack Clark confirmed the company briefed the Trump administration on Mythos, their latest AI model. The timing makes this particularly strange: the briefing happened before Trump declared an end to their working relationship. Yet according to new court filings, the Pentagon told Anthropic just a week after Trump’s declaration that the two sides were “nearly aligned.”
So what’s actually going on here?
The Mixed Signals Problem
For those of us trying to understand how AI companies work with government, this situation reveals something important: there’s often a massive gap between political announcements and actual operational reality.
Anthropic took the time to brief the Trump administration on Mythos, which the company describes as having powerful capabilities. This isn’t a casual coffee chat. Briefings like this require preparation, coordination, and usually signal an ongoing relationship. You don’t brief an administration you think is about to cut ties with you.
Then Trump makes his public declaration ending the relationship. Case closed, right?
Not quite. The Pentagon apparently didn’t get that memo, or chose to interpret it differently. Their message to Anthropic about being “nearly aligned” suggests the operational folks see things very differently from the political messaging.
Why This Matters for AI Development
This isn’t just Washington drama. When AI companies build advanced models, they face real questions about who gets briefed, when, and why. These decisions shape how AI technology develops and deploys.
Anthropic clearly felt the Trump administration needed to know about Mythos before any public announcement. That suggests either the model has capabilities that touch on national security, or Anthropic wanted government input before moving forward. Both possibilities are significant.
But here’s where it gets tricky for companies like Anthropic: if political winds can shift your government relationships overnight, how do you plan long-term development? How do you make commitments about safety testing or deployment timelines when your primary government contact might declare you persona non grata on Twitter?
The Real Story Behind the Headlines
Jack Clark’s confirmation at the Semafor World Economy summit this week adds another layer. He explained why Anthropic remained engaged despite the public turbulence. This suggests the company sees value in maintaining these connections even when the political theater gets messy.
For those of us watching the AI space, this reveals an important truth: the relationship between AI companies and government is more durable than any single announcement suggests. The technical folks keep talking to each other. The briefings continue. The work goes on.
This makes sense when you think about it. The Pentagon needs to understand what AI capabilities exist and are coming. AI companies need to understand government requirements and concerns. Neither side can afford to let political statements completely derail those conversations.
What Happens Next
The court filing revealing the Pentagon’s “nearly aligned” message suggests this story isn’t over. There’s clearly a disconnect between what Trump declared publicly and what the defense establishment thinks is happening.
For Anthropic, this creates an interesting challenge. Do you take the political declaration at face value and pull back? Or do you listen to the operational signals from the Pentagon and keep engaging?
Based on Clark’s comments about remaining engaged, Anthropic seems to be choosing the latter. That’s probably the right call for an AI company that needs to maintain relationships across administrations and political cycles.
But it does leave us with an odd situation where an AI company briefed an administration on a powerful new model, got publicly dumped, then got told by the Pentagon that everything’s actually fine. If you’re confused, you’re not alone.
This is what happens when latest technology meets political reality. The technology keeps advancing, the briefings keep happening, and the public statements don’t always match what’s going on behind closed doors.
🕒 Published: